Hi Pierre, Thanks a lot for your reply. > Your attached file didn't come through. Hmm, ok. I uploaded the wireshark-log to: http://93.190.88.182/psql_large_query.bin > - different execution plan between your app and ssh+psql, which can happen > if the planning uses/doesn't use your specific parameters, Its both times (ssh and remote psql) exactly the same query - I copied the SQL generated by hibernate and executed it in psql. And although it has many columns (~210) the result-set is only about 5 rows and am sure not larger than a few kb. > - dumb client versus smart client : > smart client : use the protocol which sends the query text + parameters + > prepare + execute in 1 TCP message, 1 ping, postgres works, 1 ping, get > reply So are both psql and the jdbc driver dumb clients? Or are there only buffers somewhere too small and therefor data is sent in many smal batches. I thought one query would more or less equal to one roundtrip, right? Maybe I should ask on the pgsql-jdbc list. > If you want to ensure the fastest response time you need to ensure than one > user action (click) needs one and only one roundtrip to the server before > all the results are displayed > One solution could be to put the database handling stuff inside an > appserver, make your app communicate to it with a low-overhead RPC protocol > (ie, not raw uncompressed XML) that minimizes the number of roudtrips, and > compresses data thoroughly. I use well tuned hibernate fetch profiles to ensure fewest possible roundtrips, however I am not getting paid well enough to create an appserver tier ;) Thanks, Clemens -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
